Only upsert once for each distinct character group

[j-muldoon]

When .wdb.writedownmode is set to `partbyfirstchar, there was a writedown happening for each unique sym, but the intended action was a writedown for each unique first character of syms.

This commit addresses this by grouping the syms by first letter, then altering the where clause in the functional select to select the corresponding syms before upserting to disk.

The number of upserts now equals the number of distinct starting characters.

[jonathonmcmurray]

I think this results in a nested list of symbols, if we can pass a simple list of symbols I think we don't need the changes to upserttopartition so maybe:

Suggested change

	if[writedownmode~`partbyfirstchar;
	extrapartitions:value extrapartitions group .Q.fu[{first each string x}; raze extrapartitions]];
	if[writedownmode~`partbyfirstchar;
	extrapartitions:raze extrapartitions;
	extrapartitions:value extrapartitions group .Q.fu[{first each string x};extrapartitions]];

or we make an alternative to .merge.getextrapartitions that returns a list instead of nested list & then we don't have to raze, e.g. something like (untested):

.merge.getfirstcharpartitions:{[tablename;extrapartitiontype]
  value ?[tablename;();({first each string x};extrapartitiontype);(distinct;extrapartitiontype)]
 }

and then we can just do something like

extrapartitions:$[writedownmode~`partbyfirstchar;.merge.getfirstcharpartitions;.merge.getextrapartitions][tablename;extrapartitiontype];

and we shouldn't have to muck around with the extrapartitions variable afterwards

[jonathonmcmurray]

Looks good - if the other alternative works we can merge as a separate PR later